linux_dsm_epyc7002/fs/nfs/direct.c

996 lines
26 KiB
C
Raw Normal View History

// SPDX-License-Identifier: GPL-2.0-only
/*
* linux/fs/nfs/direct.c
*
* Copyright (C) 2003 by Chuck Lever <cel@netapp.com>
*
* High-performance uncached I/O for the Linux NFS client
*
* There are important applications whose performance or correctness
* depends on uncached access to file data. Database clusters
* (multiple copies of the same instance running on separate hosts)
* implement their own cache coherency protocol that subsumes file
* system cache protocols. Applications that process datasets
* considerably larger than the client's memory do not always benefit
* from a local cache. A streaming video server, for instance, has no
* need to cache the contents of a file.
*
* When an application requests uncached I/O, all read and write requests
* are made directly to the server; data stored or fetched via these
* requests is not cached in the Linux page cache. The client does not
* correct unaligned requests from applications. All requested bytes are
* held on permanent storage before a direct write system call returns to
* an application.
*
* Solaris implements an uncached I/O facility called directio() that
* is used for backups and sequential I/O to very large files. Solaris
* also supports uncaching whole NFS partitions with "-o forcedirectio,"
* an undocumented mount option.
*
* Designed by Jeff Kimmel, Chuck Lever, and Trond Myklebust, with
* help from Andrew Morton.
*
* 18 Dec 2001 Initial implementation for 2.4 --cel
* 08 Jul 2002 Version for 2.4.19, with bug fixes --trondmy
* 08 Jun 2003 Port to 2.5 APIs --cel
* 31 Mar 2004 Handle direct I/O without VFS support --cel
* 15 Sep 2004 Parallel async reads --cel
* 04 May 2005 support O_DIRECT with aio --cel
*
*/
#include <linux/errno.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/file.h>
#include <linux/pagemap.h>
#include <linux/kref.h>
include cleanup: Update gfp.h and slab.h includes to prepare for breaking implicit slab.h inclusion from percpu.h percpu.h is included by sched.h and module.h and thus ends up being included when building most .c files. percpu.h includes slab.h which in turn includes gfp.h making everything defined by the two files universally available and complicating inclusion dependencies. percpu.h -> slab.h dependency is about to be removed. Prepare for this change by updating users of gfp and slab facilities include those headers directly instead of assuming availability. As this conversion needs to touch large number of source files, the following script is used as the basis of conversion. http://userweb.kernel.org/~tj/misc/slabh-sweep.py The script does the followings. * Scan files for gfp and slab usages and update includes such that only the necessary includes are there. ie. if only gfp is used, gfp.h, if slab is used, slab.h. * When the script inserts a new include, it looks at the include blocks and try to put the new include such that its order conforms to its surrounding. It's put in the include block which contains core kernel includes, in the same order that the rest are ordered - alphabetical, Christmas tree, rev-Xmas-tree or at the end if there doesn't seem to be any matching order. * If the script can't find a place to put a new include (mostly because the file doesn't have fitting include block), it prints out an error message indicating which .h file needs to be added to the file. The conversion was done in the following steps. 1. The initial automatic conversion of all .c files updated slightly over 4000 files, deleting around 700 includes and adding ~480 gfp.h and ~3000 slab.h inclusions. The script emitted errors for ~400 files. 2. Each error was manually checked. Some didn't need the inclusion, some needed manual addition while adding it to implementation .h or embedding .c file was more appropriate for others. This step added inclusions to around 150 files. 3. The script was run again and the output was compared to the edits from #2 to make sure no file was left behind. 4. Several build tests were done and a couple of problems were fixed. e.g. lib/decompress_*.c used malloc/free() wrappers around slab APIs requiring slab.h to be added manually. 5. The script was run on all .h files but without automatically editing them as sprinkling gfp.h and slab.h inclusions around .h files could easily lead to inclusion dependency hell. Most gfp.h inclusion directives were ignored as stuff from gfp.h was usually wildly available and often used in preprocessor macros. Each slab.h inclusion directive was examined and added manually as necessary. 6. percpu.h was updated not to include slab.h. 7. Build test were done on the following configurations and failures were fixed. CONFIG_GCOV_KERNEL was turned off for all tests (as my distributed build env didn't work with gcov compiles) and a few more options had to be turned off depending on archs to make things build (like ipr on powerpc/64 which failed due to missing writeq). * x86 and x86_64 UP and SMP allmodconfig and a custom test config. * powerpc and powerpc64 SMP allmodconfig * sparc and sparc64 SMP allmodconfig * ia64 SMP allmodconfig * s390 SMP allmodconfig * alpha SMP allmodconfig * um on x86_64 SMP allmodconfig 8. percpu.h modifications were reverted so that it could be applied as a separate patch and serve as bisection point. Given the fact that I had only a couple of failures from tests on step 6, I'm fairly confident about the coverage of this conversion patch. If there is a breakage, it's likely to be something in one of the arch headers which should be easily discoverable easily on most builds of the specific arch. Signed-off-by: Tejun Heo <tj@kernel.org> Guess-its-ok-by: Christoph Lameter <cl@linux-foundation.org> Cc: Ingo Molnar <mingo@redhat.com> Cc: Lee Schermerhorn <Lee.Schermerhorn@hp.com>
2010-03-24 15:04:11 +07:00
#include <linux/slab.h>
#include <linux/task_io_accounting_ops.h>
#include <linux/module.h>
#include <linux/nfs_fs.h>
#include <linux/nfs_page.h>
#include <linux/sunrpc/clnt.h>
#include <linux/uaccess.h>
#include <linux/atomic.h>
#include "internal.h"
#include "iostat.h"
#include "pnfs.h"
#define NFSDBG_FACILITY NFSDBG_VFS
static struct kmem_cache *nfs_direct_cachep;
struct nfs_direct_req {
struct kref kref; /* release manager */
/* I/O parameters */
struct nfs_open_context *ctx; /* file open context info */
struct nfs_lock_context *l_ctx; /* Lock context info */
struct kiocb * iocb; /* controlling i/o request */
struct inode * inode; /* target file of i/o */
/* completion state */
atomic_t io_count; /* i/os we're waiting for */
spinlock_t lock; /* protect completion state */
loff_t io_start; /* Start offset for I/O */
ssize_t count, /* bytes actually processed */
max_count, /* max expected count */
bytes_left, /* bytes left to be sent */
error; /* any reported error */
struct completion completion; /* wait for i/o completion */
/* commit state */
struct nfs_mds_commit_info mds_cinfo; /* Storage for cinfo */
struct pnfs_ds_commit_info ds_cinfo; /* Storage for cinfo */
struct work_struct work;
int flags;
nfs: don't dirty kernel pages read by direct-io When we use direct_IO with an NFS backing store, we can trigger a WARNING in __set_page_dirty(), as below, since we're dirtying the page unnecessarily in nfs_direct_read_completion(). To fix, replicate the logic in commit 53cbf3b157a0 ("fs: direct-io: don't dirtying pages for ITER_BVEC/ITER_KVEC direct read"). Other filesystems that implement direct_IO handle this; most use blockdev_direct_IO(). ceph and cifs have similar logic. mount 127.0.0.1:/export /nfs dd if=/dev/zero of=/nfs/image bs=1M count=200 losetup --direct-io=on -f /nfs/image mkfs.btrfs /dev/loop0 mount -t btrfs /dev/loop0 /mnt/ kernel: WARNING: CPU: 0 PID: 8067 at fs/buffer.c:580 __set_page_dirty+0xaf/0xd0 kernel: Modules linked in: loop(E) nfsv3(E) rpcsec_gss_krb5(E) nfsv4(E) dns_resolver(E) nfs(E) fscache(E) nfsd(E) auth_rpcgss(E) nfs_acl(E) lockd(E) grace(E) fuse(E) tun(E) ip6t_rpfilter(E) ipt_REJECT(E) nf_ kernel: snd_seq(E) snd_seq_device(E) snd_pcm(E) video(E) snd_timer(E) snd(E) soundcore(E) ip_tables(E) xfs(E) libcrc32c(E) sd_mod(E) sr_mod(E) cdrom(E) ata_generic(E) pata_acpi(E) crc32c_intel(E) ahci(E) li kernel: CPU: 0 PID: 8067 Comm: kworker/0:2 Tainted: G E 4.20.0-rc1.master.20181111.ol7.x86_64 #1 kernel: Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 kernel: Workqueue: nfsiod rpc_async_release [sunrpc] kernel: RIP: 0010:__set_page_dirty+0xaf/0xd0 kernel: Code: c3 48 8b 02 f6 c4 04 74 d4 48 89 df e8 ba 05 f7 ff 48 89 c6 eb cb 48 8b 43 08 a8 01 75 1f 48 89 d8 48 8b 00 a8 04 74 02 eb 87 <0f> 0b eb 83 48 83 e8 01 eb 9f 48 83 ea 01 0f 1f 00 eb 8b 48 83 e8 kernel: RSP: 0000:ffffc1c8825b7d78 EFLAGS: 00013046 kernel: RAX: 000fffffc0020089 RBX: fffff2b603308b80 RCX: 0000000000000001 kernel: RDX: 0000000000000001 RSI: ffff9d11478115c8 RDI: ffff9d11478115d0 kernel: RBP: ffffc1c8825b7da0 R08: 0000646f6973666e R09: 8080808080808080 kernel: R10: 0000000000000001 R11: 0000000000000000 R12: ffff9d11478115d0 kernel: R13: ffff9d11478115c8 R14: 0000000000003246 R15: 0000000000000001 kernel: FS: 0000000000000000(0000) GS:ffff9d115ba00000(0000) knlGS:0000000000000000 kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 kernel: CR2: 00007f408686f640 CR3: 0000000104d8e004 CR4: 00000000000606f0 kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 kernel: DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 kernel: Call Trace: kernel: __set_page_dirty_buffers+0xb6/0x110 kernel: set_page_dirty+0x52/0xb0 kernel: nfs_direct_read_completion+0xc4/0x120 [nfs] kernel: nfs_pgio_release+0x10/0x20 [nfs] kernel: rpc_free_task+0x30/0x70 [sunrpc] kernel: rpc_async_release+0x12/0x20 [sunrpc] kernel: process_one_work+0x174/0x390 kernel: worker_thread+0x4f/0x3e0 kernel: kthread+0x102/0x140 kernel: ? drain_workqueue+0x130/0x130 kernel: ? kthread_stop+0x110/0x110 kernel: ret_from_fork+0x35/0x40 kernel: ---[ end trace 01341980905412c9 ]--- Signed-off-by: Dave Kleikamp <dave.kleikamp@oracle.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com> [forward-ported to v4.20] Signed-off-by: Calum Mackay <calum.mackay@oracle.com> Reviewed-by: Dave Kleikamp <dave.kleikamp@oracle.com> Reviewed-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
2018-11-28 02:31:30 +07:00
/* for write */
#define NFS_ODIRECT_DO_COMMIT (1) /* an unstable reply was received */
#define NFS_ODIRECT_RESCHED_WRITES (2) /* write verification failed */
nfs: don't dirty kernel pages read by direct-io When we use direct_IO with an NFS backing store, we can trigger a WARNING in __set_page_dirty(), as below, since we're dirtying the page unnecessarily in nfs_direct_read_completion(). To fix, replicate the logic in commit 53cbf3b157a0 ("fs: direct-io: don't dirtying pages for ITER_BVEC/ITER_KVEC direct read"). Other filesystems that implement direct_IO handle this; most use blockdev_direct_IO(). ceph and cifs have similar logic. mount 127.0.0.1:/export /nfs dd if=/dev/zero of=/nfs/image bs=1M count=200 losetup --direct-io=on -f /nfs/image mkfs.btrfs /dev/loop0 mount -t btrfs /dev/loop0 /mnt/ kernel: WARNING: CPU: 0 PID: 8067 at fs/buffer.c:580 __set_page_dirty+0xaf/0xd0 kernel: Modules linked in: loop(E) nfsv3(E) rpcsec_gss_krb5(E) nfsv4(E) dns_resolver(E) nfs(E) fscache(E) nfsd(E) auth_rpcgss(E) nfs_acl(E) lockd(E) grace(E) fuse(E) tun(E) ip6t_rpfilter(E) ipt_REJECT(E) nf_ kernel: snd_seq(E) snd_seq_device(E) snd_pcm(E) video(E) snd_timer(E) snd(E) soundcore(E) ip_tables(E) xfs(E) libcrc32c(E) sd_mod(E) sr_mod(E) cdrom(E) ata_generic(E) pata_acpi(E) crc32c_intel(E) ahci(E) li kernel: CPU: 0 PID: 8067 Comm: kworker/0:2 Tainted: G E 4.20.0-rc1.master.20181111.ol7.x86_64 #1 kernel: Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 kernel: Workqueue: nfsiod rpc_async_release [sunrpc] kernel: RIP: 0010:__set_page_dirty+0xaf/0xd0 kernel: Code: c3 48 8b 02 f6 c4 04 74 d4 48 89 df e8 ba 05 f7 ff 48 89 c6 eb cb 48 8b 43 08 a8 01 75 1f 48 89 d8 48 8b 00 a8 04 74 02 eb 87 <0f> 0b eb 83 48 83 e8 01 eb 9f 48 83 ea 01 0f 1f 00 eb 8b 48 83 e8 kernel: RSP: 0000:ffffc1c8825b7d78 EFLAGS: 00013046 kernel: RAX: 000fffffc0020089 RBX: fffff2b603308b80 RCX: 0000000000000001 kernel: RDX: 0000000000000001 RSI: ffff9d11478115c8 RDI: ffff9d11478115d0 kernel: RBP: ffffc1c8825b7da0 R08: 0000646f6973666e R09: 8080808080808080 kernel: R10: 0000000000000001 R11: 0000000000000000 R12: ffff9d11478115d0 kernel: R13: ffff9d11478115c8 R14: 0000000000003246 R15: 0000000000000001 kernel: FS: 0000000000000000(0000) GS:ffff9d115ba00000(0000) knlGS:0000000000000000 kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 kernel: CR2: 00007f408686f640 CR3: 0000000104d8e004 CR4: 00000000000606f0 kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 kernel: DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 kernel: Call Trace: kernel: __set_page_dirty_buffers+0xb6/0x110 kernel: set_page_dirty+0x52/0xb0 kernel: nfs_direct_read_completion+0xc4/0x120 [nfs] kernel: nfs_pgio_release+0x10/0x20 [nfs] kernel: rpc_free_task+0x30/0x70 [sunrpc] kernel: rpc_async_release+0x12/0x20 [sunrpc] kernel: process_one_work+0x174/0x390 kernel: worker_thread+0x4f/0x3e0 kernel: kthread+0x102/0x140 kernel: ? drain_workqueue+0x130/0x130 kernel: ? kthread_stop+0x110/0x110 kernel: ret_from_fork+0x35/0x40 kernel: ---[ end trace 01341980905412c9 ]--- Signed-off-by: Dave Kleikamp <dave.kleikamp@oracle.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com> [forward-ported to v4.20] Signed-off-by: Calum Mackay <calum.mackay@oracle.com> Reviewed-by: Dave Kleikamp <dave.kleikamp@oracle.com> Reviewed-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
2018-11-28 02:31:30 +07:00
/* for read */
#define NFS_ODIRECT_SHOULD_DIRTY (3) /* dirty user-space page after read */
#define NFS_ODIRECT_DONE INT_MAX /* write verification failed */
};
static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops;
static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops;
static void nfs_direct_write_complete(struct nfs_direct_req *dreq);
static void nfs_direct_write_schedule_work(struct work_struct *work);
static inline void get_dreq(struct nfs_direct_req *dreq)
{
atomic_inc(&dreq->io_count);
}
static inline int put_dreq(struct nfs_direct_req *dreq)
{
return atomic_dec_and_test(&dreq->io_count);
}
static void
nfs_direct_handle_truncated(struct nfs_direct_req *dreq,
const struct nfs_pgio_header *hdr,
ssize_t dreq_len)
{
if (!(test_bit(NFS_IOHDR_ERROR, &hdr->flags) ||
test_bit(NFS_IOHDR_EOF, &hdr->flags)))
return;
if (dreq->max_count >= dreq_len) {
dreq->max_count = dreq_len;
if (dreq->count > dreq_len)
dreq->count = dreq_len;
if (test_bit(NFS_IOHDR_ERROR, &hdr->flags))
dreq->error = hdr->error;
else /* Clear outstanding error if this is EOF */
dreq->error = 0;
}
}
static void
nfs_direct_count_bytes(struct nfs_direct_req *dreq,
const struct nfs_pgio_header *hdr)
{
loff_t hdr_end = hdr->io_start + hdr->good_bytes;
ssize_t dreq_len = 0;
if (hdr_end > dreq->io_start)
dreq_len = hdr_end - dreq->io_start;
nfs_direct_handle_truncated(dreq, hdr, dreq_len);
if (dreq_len > dreq->max_count)
dreq_len = dreq->max_count;
if (dreq->count < dreq_len)
dreq->count = dreq_len;
}
/**
* nfs_direct_IO - NFS address space operation for direct I/O
* @iocb: target I/O control block
* @iter: I/O buffer
*
* The presence of this routine in the address space ops vector means
* the NFS client supports direct I/O. However, for most direct IO, we
* shunt off direct read and write requests before the VFS gets them,
* so this method is only ever called for swap.
*/
ssize_t nfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
{
struct inode *inode = iocb->ki_filp->f_mapping->host;
/* we only support swap file calling nfs_direct_IO */
if (!IS_SWAPFILE(inode))
return 0;
VM_BUG_ON(iov_iter_count(iter) != PAGE_SIZE);
if (iov_iter_rw(iter) == READ)
return nfs_file_direct_read(iocb, iter);
return nfs_file_direct_write(iocb, iter);
}
static void nfs_direct_release_pages(struct page **pages, unsigned int npages)
{
unsigned int i;
for (i = 0; i < npages; i++)
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
put_page(pages[i]);
}
void nfs_init_cinfo_from_dreq(struct nfs_commit_info *cinfo,
struct nfs_direct_req *dreq)
{
cinfo->inode = dreq->inode;
cinfo->mds = &dreq->mds_cinfo;
cinfo->ds = &dreq->ds_cinfo;
cinfo->dreq = dreq;
cinfo->completion_ops = &nfs_direct_commit_completion_ops;
}
static inline struct nfs_direct_req *nfs_direct_req_alloc(void)
{
struct nfs_direct_req *dreq;
dreq = kmem_cache_zalloc(nfs_direct_cachep, GFP_KERNEL);
if (!dreq)
return NULL;
kref_init(&dreq->kref);
kref_get(&dreq->kref);
init_completion(&dreq->completion);
INIT_LIST_HEAD(&dreq->mds_cinfo.list);
pnfs_init_ds_commit_info(&dreq->ds_cinfo);
INIT_WORK(&dreq->work, nfs_direct_write_schedule_work);
spin_lock_init(&dreq->lock);
return dreq;
}
static void nfs_direct_req_free(struct kref *kref)
{
struct nfs_direct_req *dreq = container_of(kref, struct nfs_direct_req, kref);
pnfs_release_ds_info(&dreq->ds_cinfo, dreq->inode);
if (dreq->l_ctx != NULL)
nfs_put_lock_context(dreq->l_ctx);
if (dreq->ctx != NULL)
put_nfs_open_context(dreq->ctx);
kmem_cache_free(nfs_direct_cachep, dreq);
}
static void nfs_direct_req_release(struct nfs_direct_req *dreq)
{
kref_put(&dreq->kref, nfs_direct_req_free);
}
ssize_t nfs_dreq_bytes_left(struct nfs_direct_req *dreq)
{
return dreq->bytes_left;
}
EXPORT_SYMBOL_GPL(nfs_dreq_bytes_left);
/*
* Collects and returns the final error value/byte-count.
*/
static ssize_t nfs_direct_wait(struct nfs_direct_req *dreq)
{
ssize_t result = -EIOCBQUEUED;
/* Async requests don't wait here */
if (dreq->iocb)
goto out;
result = wait_for_completion_killable(&dreq->completion);
if (!result) {
result = dreq->count;
WARN_ON_ONCE(dreq->count < 0);
}
if (!result)
result = dreq->error;
out:
return (ssize_t) result;
}
/*
* Synchronous I/O uses a stack-allocated iocb. Thus we can't trust
* the iocb is still valid here if this is a synchronous request.
*/
static void nfs_direct_complete(struct nfs_direct_req *dreq)
{
struct inode *inode = dreq->inode;
direct-io: only inc/dec inode->i_dio_count for file systems do_blockdev_direct_IO() increments and decrements the inode ->i_dio_count for each IO operation. It does this to protect against truncate of a file. Block devices don't need this sort of protection. For a capable multiqueue setup, this atomic int is the only shared state between applications accessing the device for O_DIRECT, and it presents a scaling wall for that. In my testing, as much as 30% of system time is spent incrementing and decrementing this value. A mixed read/write workload improved from ~2.5M IOPS to ~9.6M IOPS, with better latencies too. Before: clat percentiles (usec): | 1.00th=[ 33], 5.00th=[ 34], 10.00th=[ 34], 20.00th=[ 34], | 30.00th=[ 34], 40.00th=[ 34], 50.00th=[ 35], 60.00th=[ 35], | 70.00th=[ 35], 80.00th=[ 35], 90.00th=[ 37], 95.00th=[ 80], | 99.00th=[ 98], 99.50th=[ 151], 99.90th=[ 155], 99.95th=[ 155], | 99.99th=[ 165] After: clat percentiles (usec): | 1.00th=[ 95], 5.00th=[ 108], 10.00th=[ 129], 20.00th=[ 149], | 30.00th=[ 155], 40.00th=[ 161], 50.00th=[ 167], 60.00th=[ 171], | 70.00th=[ 177], 80.00th=[ 185], 90.00th=[ 201], 95.00th=[ 270], | 99.00th=[ 390], 99.50th=[ 398], 99.90th=[ 418], 99.95th=[ 422], | 99.99th=[ 438] In other setups, Robert Elliott reported seeing good performance improvements: https://lkml.org/lkml/2015/4/3/557 The more applications accessing the device, the worse it gets. Add a new direct-io flags, DIO_SKIP_DIO_COUNT, which tells do_blockdev_direct_IO() that it need not worry about incrementing or decrementing the inode i_dio_count for this caller. Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Elliott, Robert (Server Storage) <elliott@hp.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2015-04-16 06:05:48 +07:00
inode_dio_end(inode);
if (dreq->iocb) {
long res = (long) dreq->error;
if (dreq->count != 0) {
res = (long) dreq->count;
WARN_ON_ONCE(dreq->count < 0);
}
dreq->iocb->ki_complete(dreq->iocb, res, 0);
}
NFS: direct: use complete() instead of complete_all() There is only one waiter for the completion, therefore there is no need to use complete_all(). Let's make that clear by using complete() instead of complete_all(). nfs_file_direct_write() or nfs_file_direct_read() allocated a request object via nfs_direct_req_alloc(), which initializes the completion. The request object then is freed later in the exit path. Between the initialization and the release either nfs_direct_write_schedule_iovec() resp nfs_direct_read_schedule_iovec() are called which will asynchronously process the request. The calling function waits via nfs_direct_wait() till the async work has been done. Thus there is only one waiter on the completion. nfs_direct_pgio_init() and nfs_direct_read_completion() are passed via function pointers to nfs pageio. The first function does a ref counting (get_dreq() and put_dreq()) which ensures that nfs_direct_read_completion() and nfs_direct_read_schedule_iovec() only call the completion path once. The usage pattern of the completion is: waiter context waker context nfs_file_direct_write() dreq = nfs_direct_req_alloc() init_completion() nfs_direct_write_schedule_iovec() nfs_direct_wait() wait_for_completion_killable() nfs_direct_write_schedule_work() nfs_direct_complete() complete() nfs_file_direct_read() dreq = nfs_direct_req_all() init_completion() nfs_direct_read_schedule_iovec() nfs_direct_wait() wait_for_completion_killable() nfs_direct_read_schedule_iovec() nfs_direct_complete() complete() nfs_direct_read_completion() nfs_direct_complete() complete() Signed-off-by: Daniel Wagner <daniel.wagner@bmw-carit.de> Signed-off-by: Anna Schumaker <Anna.Schumaker@Netapp.com>
2016-09-22 18:54:28 +07:00
complete(&dreq->completion);
nfs_direct_req_release(dreq);
}
static void nfs_direct_read_completion(struct nfs_pgio_header *hdr)
{
unsigned long bytes = 0;
struct nfs_direct_req *dreq = hdr->dreq;
spin_lock(&dreq->lock);
if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) {
spin_unlock(&dreq->lock);
goto out_put;
}
nfs_direct_count_bytes(dreq, hdr);
spin_unlock(&dreq->lock);
while (!list_empty(&hdr->pages)) {
struct nfs_page *req = nfs_list_entry(hdr->pages.next);
struct page *page = req->wb_page;
nfs: don't dirty kernel pages read by direct-io When we use direct_IO with an NFS backing store, we can trigger a WARNING in __set_page_dirty(), as below, since we're dirtying the page unnecessarily in nfs_direct_read_completion(). To fix, replicate the logic in commit 53cbf3b157a0 ("fs: direct-io: don't dirtying pages for ITER_BVEC/ITER_KVEC direct read"). Other filesystems that implement direct_IO handle this; most use blockdev_direct_IO(). ceph and cifs have similar logic. mount 127.0.0.1:/export /nfs dd if=/dev/zero of=/nfs/image bs=1M count=200 losetup --direct-io=on -f /nfs/image mkfs.btrfs /dev/loop0 mount -t btrfs /dev/loop0 /mnt/ kernel: WARNING: CPU: 0 PID: 8067 at fs/buffer.c:580 __set_page_dirty+0xaf/0xd0 kernel: Modules linked in: loop(E) nfsv3(E) rpcsec_gss_krb5(E) nfsv4(E) dns_resolver(E) nfs(E) fscache(E) nfsd(E) auth_rpcgss(E) nfs_acl(E) lockd(E) grace(E) fuse(E) tun(E) ip6t_rpfilter(E) ipt_REJECT(E) nf_ kernel: snd_seq(E) snd_seq_device(E) snd_pcm(E) video(E) snd_timer(E) snd(E) soundcore(E) ip_tables(E) xfs(E) libcrc32c(E) sd_mod(E) sr_mod(E) cdrom(E) ata_generic(E) pata_acpi(E) crc32c_intel(E) ahci(E) li kernel: CPU: 0 PID: 8067 Comm: kworker/0:2 Tainted: G E 4.20.0-rc1.master.20181111.ol7.x86_64 #1 kernel: Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 kernel: Workqueue: nfsiod rpc_async_release [sunrpc] kernel: RIP: 0010:__set_page_dirty+0xaf/0xd0 kernel: Code: c3 48 8b 02 f6 c4 04 74 d4 48 89 df e8 ba 05 f7 ff 48 89 c6 eb cb 48 8b 43 08 a8 01 75 1f 48 89 d8 48 8b 00 a8 04 74 02 eb 87 <0f> 0b eb 83 48 83 e8 01 eb 9f 48 83 ea 01 0f 1f 00 eb 8b 48 83 e8 kernel: RSP: 0000:ffffc1c8825b7d78 EFLAGS: 00013046 kernel: RAX: 000fffffc0020089 RBX: fffff2b603308b80 RCX: 0000000000000001 kernel: RDX: 0000000000000001 RSI: ffff9d11478115c8 RDI: ffff9d11478115d0 kernel: RBP: ffffc1c8825b7da0 R08: 0000646f6973666e R09: 8080808080808080 kernel: R10: 0000000000000001 R11: 0000000000000000 R12: ffff9d11478115d0 kernel: R13: ffff9d11478115c8 R14: 0000000000003246 R15: 0000000000000001 kernel: FS: 0000000000000000(0000) GS:ffff9d115ba00000(0000) knlGS:0000000000000000 kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 kernel: CR2: 00007f408686f640 CR3: 0000000104d8e004 CR4: 00000000000606f0 kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 kernel: DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 kernel: Call Trace: kernel: __set_page_dirty_buffers+0xb6/0x110 kernel: set_page_dirty+0x52/0xb0 kernel: nfs_direct_read_completion+0xc4/0x120 [nfs] kernel: nfs_pgio_release+0x10/0x20 [nfs] kernel: rpc_free_task+0x30/0x70 [sunrpc] kernel: rpc_async_release+0x12/0x20 [sunrpc] kernel: process_one_work+0x174/0x390 kernel: worker_thread+0x4f/0x3e0 kernel: kthread+0x102/0x140 kernel: ? drain_workqueue+0x130/0x130 kernel: ? kthread_stop+0x110/0x110 kernel: ret_from_fork+0x35/0x40 kernel: ---[ end trace 01341980905412c9 ]--- Signed-off-by: Dave Kleikamp <dave.kleikamp@oracle.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com> [forward-ported to v4.20] Signed-off-by: Calum Mackay <calum.mackay@oracle.com> Reviewed-by: Dave Kleikamp <dave.kleikamp@oracle.com> Reviewed-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
2018-11-28 02:31:30 +07:00
if (!PageCompound(page) && bytes < hdr->good_bytes &&
(dreq->flags == NFS_ODIRECT_SHOULD_DIRTY))
set_page_dirty(page);
bytes += req->wb_bytes;
nfs_list_remove_request(req);
nfs_release_request(req);
}
out_put:
if (put_dreq(dreq))
nfs_direct_complete(dreq);
hdr->release(hdr);
}
static void nfs_read_sync_pgio_error(struct list_head *head, int error)
{
struct nfs_page *req;
while (!list_empty(head)) {
req = nfs_list_entry(head->next);
nfs_list_remove_request(req);
nfs_release_request(req);
}
}
static void nfs_direct_pgio_init(struct nfs_pgio_header *hdr)
{
get_dreq(hdr->dreq);
}
static const struct nfs_pgio_completion_ops nfs_direct_read_completion_ops = {
.error_cleanup = nfs_read_sync_pgio_error,
.init_hdr = nfs_direct_pgio_init,
.completion = nfs_direct_read_completion,
};
/*
* For each rsize'd chunk of the user's buffer, dispatch an NFS READ
* operation. If nfs_readdata_alloc() or get_user_pages() fails,
* bail and stop sending more reads. Read length accounting is
* handled automatically by nfs_direct_read_result(). Otherwise, if
* no requests have been sent, just return an error.
*/
static ssize_t nfs_direct_read_schedule_iovec(struct nfs_direct_req *dreq,
struct iov_iter *iter,
loff_t pos)
{
struct nfs_pageio_descriptor desc;
struct inode *inode = dreq->inode;
ssize_t result = -EINVAL;
size_t requested_bytes = 0;
size_t rsize = max_t(size_t, NFS_SERVER(inode)->rsize, PAGE_SIZE);
2014-06-13 00:30:18 +07:00
nfs_pageio_init_read(&desc, dreq->inode, false,
&nfs_direct_read_completion_ops);
get_dreq(dreq);
desc.pg_dreq = dreq;
direct-io: only inc/dec inode->i_dio_count for file systems do_blockdev_direct_IO() increments and decrements the inode ->i_dio_count for each IO operation. It does this to protect against truncate of a file. Block devices don't need this sort of protection. For a capable multiqueue setup, this atomic int is the only shared state between applications accessing the device for O_DIRECT, and it presents a scaling wall for that. In my testing, as much as 30% of system time is spent incrementing and decrementing this value. A mixed read/write workload improved from ~2.5M IOPS to ~9.6M IOPS, with better latencies too. Before: clat percentiles (usec): | 1.00th=[ 33], 5.00th=[ 34], 10.00th=[ 34], 20.00th=[ 34], | 30.00th=[ 34], 40.00th=[ 34], 50.00th=[ 35], 60.00th=[ 35], | 70.00th=[ 35], 80.00th=[ 35], 90.00th=[ 37], 95.00th=[ 80], | 99.00th=[ 98], 99.50th=[ 151], 99.90th=[ 155], 99.95th=[ 155], | 99.99th=[ 165] After: clat percentiles (usec): | 1.00th=[ 95], 5.00th=[ 108], 10.00th=[ 129], 20.00th=[ 149], | 30.00th=[ 155], 40.00th=[ 161], 50.00th=[ 167], 60.00th=[ 171], | 70.00th=[ 177], 80.00th=[ 185], 90.00th=[ 201], 95.00th=[ 270], | 99.00th=[ 390], 99.50th=[ 398], 99.90th=[ 418], 99.95th=[ 422], | 99.99th=[ 438] In other setups, Robert Elliott reported seeing good performance improvements: https://lkml.org/lkml/2015/4/3/557 The more applications accessing the device, the worse it gets. Add a new direct-io flags, DIO_SKIP_DIO_COUNT, which tells do_blockdev_direct_IO() that it need not worry about incrementing or decrementing the inode i_dio_count for this caller. Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Elliott, Robert (Server Storage) <elliott@hp.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2015-04-16 06:05:48 +07:00
inode_dio_begin(inode);
while (iov_iter_count(iter)) {
struct page **pagevec;
size_t bytes;
size_t pgbase;
unsigned npages, i;
result = iov_iter_get_pages_alloc(iter, &pagevec,
rsize, &pgbase);
if (result < 0)
break;
bytes = result;
iov_iter_advance(iter, bytes);
npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE;
for (i = 0; i < npages; i++) {
struct nfs_page *req;
unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
/* XXX do we need to do the eof zeroing found in async_filler? */
req = nfs_create_request(dreq->ctx, pagevec[i],
pgbase, req_len);
if (IS_ERR(req)) {
result = PTR_ERR(req);
break;
}
req->wb_index = pos >> PAGE_SHIFT;
req->wb_offset = pos & ~PAGE_MASK;
if (!nfs_pageio_add_request(&desc, req)) {
result = desc.pg_error;
nfs_release_request(req);
break;
}
pgbase = 0;
bytes -= req_len;
requested_bytes += req_len;
pos += req_len;
dreq->bytes_left -= req_len;
}
nfs_direct_release_pages(pagevec, npages);
kvfree(pagevec);
if (result < 0)
break;
}
nfs_pageio_complete(&desc);
NFS: Fix "kernel BUG at fs/aio.c:554!" Nick Piggin reports: > I'm getting use after frees in aio code in NFS > > [ 2703.396766] Call Trace: > [ 2703.396858] [<ffffffff8100b057>] ? native_sched_clock+0x27/0x80 > [ 2703.396959] [<ffffffff8108509e>] ? put_lock_stats+0xe/0x40 > [ 2703.397058] [<ffffffff81088348>] ? lock_release_holdtime+0xa8/0x140 > [ 2703.397159] [<ffffffff8108a2a5>] lock_acquire+0x95/0x1b0 > [ 2703.397260] [<ffffffff811627db>] ? aio_put_req+0x2b/0x60 > [ 2703.397361] [<ffffffff81039701>] ? get_parent_ip+0x11/0x50 > [ 2703.397464] [<ffffffff81612a31>] _raw_spin_lock_irq+0x41/0x80 > [ 2703.397564] [<ffffffff811627db>] ? aio_put_req+0x2b/0x60 > [ 2703.397662] [<ffffffff811627db>] aio_put_req+0x2b/0x60 > [ 2703.397761] [<ffffffff811647fe>] do_io_submit+0x2be/0x7c0 > [ 2703.397895] [<ffffffff81164d0b>] sys_io_submit+0xb/0x10 > [ 2703.397995] [<ffffffff8100307b>] system_call_fastpath+0x16/0x1b > > Adding some tracing, it is due to nfs completing the request then > returning something other than -EIOCBQUEUED, so aio.c > also completes the request. To address this, prevent the NFS direct I/O engine from completing async iocbs when the forward path returns an error without starting any I/O. This fix appears to survive ^C during both "xfstest no. 208" and "fsx -Z." It's likely this bug has existed for a very long while, as we are seeing very similar symptoms in OEL 5. Copying stable. Cc: Stable <stable@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-01-21 22:54:57 +07:00
/*
* If no bytes were started, return the error, and let the
* generic layer handle the completion.
*/
if (requested_bytes == 0) {
direct-io: only inc/dec inode->i_dio_count for file systems do_blockdev_direct_IO() increments and decrements the inode ->i_dio_count for each IO operation. It does this to protect against truncate of a file. Block devices don't need this sort of protection. For a capable multiqueue setup, this atomic int is the only shared state between applications accessing the device for O_DIRECT, and it presents a scaling wall for that. In my testing, as much as 30% of system time is spent incrementing and decrementing this value. A mixed read/write workload improved from ~2.5M IOPS to ~9.6M IOPS, with better latencies too. Before: clat percentiles (usec): | 1.00th=[ 33], 5.00th=[ 34], 10.00th=[ 34], 20.00th=[ 34], | 30.00th=[ 34], 40.00th=[ 34], 50.00th=[ 35], 60.00th=[ 35], | 70.00th=[ 35], 80.00th=[ 35], 90.00th=[ 37], 95.00th=[ 80], | 99.00th=[ 98], 99.50th=[ 151], 99.90th=[ 155], 99.95th=[ 155], | 99.99th=[ 165] After: clat percentiles (usec): | 1.00th=[ 95], 5.00th=[ 108], 10.00th=[ 129], 20.00th=[ 149], | 30.00th=[ 155], 40.00th=[ 161], 50.00th=[ 167], 60.00th=[ 171], | 70.00th=[ 177], 80.00th=[ 185], 90.00th=[ 201], 95.00th=[ 270], | 99.00th=[ 390], 99.50th=[ 398], 99.90th=[ 418], 99.95th=[ 422], | 99.99th=[ 438] In other setups, Robert Elliott reported seeing good performance improvements: https://lkml.org/lkml/2015/4/3/557 The more applications accessing the device, the worse it gets. Add a new direct-io flags, DIO_SKIP_DIO_COUNT, which tells do_blockdev_direct_IO() that it need not worry about incrementing or decrementing the inode i_dio_count for this caller. Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Elliott, Robert (Server Storage) <elliott@hp.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2015-04-16 06:05:48 +07:00
inode_dio_end(inode);
NFS: Fix "kernel BUG at fs/aio.c:554!" Nick Piggin reports: > I'm getting use after frees in aio code in NFS > > [ 2703.396766] Call Trace: > [ 2703.396858] [<ffffffff8100b057>] ? native_sched_clock+0x27/0x80 > [ 2703.396959] [<ffffffff8108509e>] ? put_lock_stats+0xe/0x40 > [ 2703.397058] [<ffffffff81088348>] ? lock_release_holdtime+0xa8/0x140 > [ 2703.397159] [<ffffffff8108a2a5>] lock_acquire+0x95/0x1b0 > [ 2703.397260] [<ffffffff811627db>] ? aio_put_req+0x2b/0x60 > [ 2703.397361] [<ffffffff81039701>] ? get_parent_ip+0x11/0x50 > [ 2703.397464] [<ffffffff81612a31>] _raw_spin_lock_irq+0x41/0x80 > [ 2703.397564] [<ffffffff811627db>] ? aio_put_req+0x2b/0x60 > [ 2703.397662] [<ffffffff811627db>] aio_put_req+0x2b/0x60 > [ 2703.397761] [<ffffffff811647fe>] do_io_submit+0x2be/0x7c0 > [ 2703.397895] [<ffffffff81164d0b>] sys_io_submit+0xb/0x10 > [ 2703.397995] [<ffffffff8100307b>] system_call_fastpath+0x16/0x1b > > Adding some tracing, it is due to nfs completing the request then > returning something other than -EIOCBQUEUED, so aio.c > also completes the request. To address this, prevent the NFS direct I/O engine from completing async iocbs when the forward path returns an error without starting any I/O. This fix appears to survive ^C during both "xfstest no. 208" and "fsx -Z." It's likely this bug has existed for a very long while, as we are seeing very similar symptoms in OEL 5. Copying stable. Cc: Stable <stable@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-01-21 22:54:57 +07:00
nfs_direct_req_release(dreq);
return result < 0 ? result : -EIO;
}
if (put_dreq(dreq))
nfs_direct_complete(dreq);
return requested_bytes;
}
/**
* nfs_file_direct_read - file direct read operation for NFS files
* @iocb: target I/O control block
* @iter: vector of user buffers into which to read data
*
* We use this function for direct reads instead of calling
* generic_file_aio_read() in order to avoid gfar's check to see if
* the request starts before the end of the file. For that check
* to work, we must generate a GETATTR before each direct read, and
* even then there is a window between the GETATTR and the subsequent
* READ where the file size could change. Our preference is simply
* to do all reads the application wants, and the server will take
* care of managing the end of file boundary.
*
* This function also eliminates unnecessarily updating the file's
* atime locally, as the NFS server sets the file's atime, and this
* client must read the updated atime from the server back into its
* cache.
*/
ssize_t nfs_file_direct_read(struct kiocb *iocb, struct iov_iter *iter)
{
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
struct nfs_direct_req *dreq;
struct nfs_lock_context *l_ctx;
ssize_t result, requested;
size_t count = iov_iter_count(iter);
nfs_add_stats(mapping->host, NFSIOS_DIRECTREADBYTES, count);
dfprintk(FILE, "NFS: direct read(%pD2, %zd@%Ld)\n",
file, count, (long long) iocb->ki_pos);
result = 0;
if (!count)
goto out;
task_io_account_read(count);
result = -ENOMEM;
dreq = nfs_direct_req_alloc();
if (dreq == NULL)
goto out;
dreq->inode = inode;
dreq->bytes_left = dreq->max_count = count;
dreq->io_start = iocb->ki_pos;
dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
l_ctx = nfs_get_lock_context(dreq->ctx);
if (IS_ERR(l_ctx)) {
result = PTR_ERR(l_ctx);
nfs_direct_req_release(dreq);
goto out_release;
}
dreq->l_ctx = l_ctx;
if (!is_sync_kiocb(iocb))
dreq->iocb = iocb;
nfs: don't dirty kernel pages read by direct-io When we use direct_IO with an NFS backing store, we can trigger a WARNING in __set_page_dirty(), as below, since we're dirtying the page unnecessarily in nfs_direct_read_completion(). To fix, replicate the logic in commit 53cbf3b157a0 ("fs: direct-io: don't dirtying pages for ITER_BVEC/ITER_KVEC direct read"). Other filesystems that implement direct_IO handle this; most use blockdev_direct_IO(). ceph and cifs have similar logic. mount 127.0.0.1:/export /nfs dd if=/dev/zero of=/nfs/image bs=1M count=200 losetup --direct-io=on -f /nfs/image mkfs.btrfs /dev/loop0 mount -t btrfs /dev/loop0 /mnt/ kernel: WARNING: CPU: 0 PID: 8067 at fs/buffer.c:580 __set_page_dirty+0xaf/0xd0 kernel: Modules linked in: loop(E) nfsv3(E) rpcsec_gss_krb5(E) nfsv4(E) dns_resolver(E) nfs(E) fscache(E) nfsd(E) auth_rpcgss(E) nfs_acl(E) lockd(E) grace(E) fuse(E) tun(E) ip6t_rpfilter(E) ipt_REJECT(E) nf_ kernel: snd_seq(E) snd_seq_device(E) snd_pcm(E) video(E) snd_timer(E) snd(E) soundcore(E) ip_tables(E) xfs(E) libcrc32c(E) sd_mod(E) sr_mod(E) cdrom(E) ata_generic(E) pata_acpi(E) crc32c_intel(E) ahci(E) li kernel: CPU: 0 PID: 8067 Comm: kworker/0:2 Tainted: G E 4.20.0-rc1.master.20181111.ol7.x86_64 #1 kernel: Hardware name: innotek GmbH VirtualBox/VirtualBox, BIOS VirtualBox 12/01/2006 kernel: Workqueue: nfsiod rpc_async_release [sunrpc] kernel: RIP: 0010:__set_page_dirty+0xaf/0xd0 kernel: Code: c3 48 8b 02 f6 c4 04 74 d4 48 89 df e8 ba 05 f7 ff 48 89 c6 eb cb 48 8b 43 08 a8 01 75 1f 48 89 d8 48 8b 00 a8 04 74 02 eb 87 <0f> 0b eb 83 48 83 e8 01 eb 9f 48 83 ea 01 0f 1f 00 eb 8b 48 83 e8 kernel: RSP: 0000:ffffc1c8825b7d78 EFLAGS: 00013046 kernel: RAX: 000fffffc0020089 RBX: fffff2b603308b80 RCX: 0000000000000001 kernel: RDX: 0000000000000001 RSI: ffff9d11478115c8 RDI: ffff9d11478115d0 kernel: RBP: ffffc1c8825b7da0 R08: 0000646f6973666e R09: 8080808080808080 kernel: R10: 0000000000000001 R11: 0000000000000000 R12: ffff9d11478115d0 kernel: R13: ffff9d11478115c8 R14: 0000000000003246 R15: 0000000000000001 kernel: FS: 0000000000000000(0000) GS:ffff9d115ba00000(0000) knlGS:0000000000000000 kernel: CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 kernel: CR2: 00007f408686f640 CR3: 0000000104d8e004 CR4: 00000000000606f0 kernel: DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 kernel: DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 kernel: Call Trace: kernel: __set_page_dirty_buffers+0xb6/0x110 kernel: set_page_dirty+0x52/0xb0 kernel: nfs_direct_read_completion+0xc4/0x120 [nfs] kernel: nfs_pgio_release+0x10/0x20 [nfs] kernel: rpc_free_task+0x30/0x70 [sunrpc] kernel: rpc_async_release+0x12/0x20 [sunrpc] kernel: process_one_work+0x174/0x390 kernel: worker_thread+0x4f/0x3e0 kernel: kthread+0x102/0x140 kernel: ? drain_workqueue+0x130/0x130 kernel: ? kthread_stop+0x110/0x110 kernel: ret_from_fork+0x35/0x40 kernel: ---[ end trace 01341980905412c9 ]--- Signed-off-by: Dave Kleikamp <dave.kleikamp@oracle.com> Signed-off-by: Santosh Shilimkar <santosh.shilimkar@oracle.com> [forward-ported to v4.20] Signed-off-by: Calum Mackay <calum.mackay@oracle.com> Reviewed-by: Dave Kleikamp <dave.kleikamp@oracle.com> Reviewed-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <trond.myklebust@hammerspace.com>
2018-11-28 02:31:30 +07:00
if (iter_is_iovec(iter))
dreq->flags = NFS_ODIRECT_SHOULD_DIRTY;
nfs_start_io_direct(inode);
NFS_I(inode)->read_io += count;
requested = nfs_direct_read_schedule_iovec(dreq, iter, iocb->ki_pos);
nfs_end_io_direct(inode);
if (requested > 0) {
result = nfs_direct_wait(dreq);
if (result > 0) {
requested -= result;
iocb->ki_pos += result;
}
iov_iter_revert(iter, requested);
} else {
result = requested;
}
out_release:
nfs_direct_req_release(dreq);
out:
return result;
}
static void
nfs_direct_join_group(struct list_head *list, struct inode *inode)
{
struct nfs_page *req, *next;
list_for_each_entry(req, list, wb_list) {
if (req->wb_head != req || req->wb_this_page == req)
continue;
for (next = req->wb_this_page;
next != req->wb_head;
next = next->wb_this_page) {
nfs_list_remove_request(next);
nfs_release_request(next);
}
nfs_join_page_group(req, inode);
}
}
static void
nfs_direct_write_scan_commit_list(struct inode *inode,
struct list_head *list,
struct nfs_commit_info *cinfo)
{
mutex_lock(&NFS_I(cinfo->inode)->commit_mutex);
pnfs_recover_commit_reqs(list, cinfo);
nfs_scan_commit_list(&cinfo->mds->list, list, cinfo, 0);
mutex_unlock(&NFS_I(cinfo->inode)->commit_mutex);
}
static void nfs_direct_write_reschedule(struct nfs_direct_req *dreq)
{
struct nfs_pageio_descriptor desc;
struct nfs_page *req, *tmp;
LIST_HEAD(reqs);
struct nfs_commit_info cinfo;
LIST_HEAD(failed);
nfs_init_cinfo_from_dreq(&cinfo, dreq);
nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo);
nfs_direct_join_group(&reqs, dreq->inode);
dreq->count = 0;
dreq->max_count = 0;
list_for_each_entry(req, &reqs, wb_list)
dreq->max_count += req->wb_bytes;
nfs_clear_pnfs_ds_commit_verifiers(&dreq->ds_cinfo);
get_dreq(dreq);
nfs_pageio_init_write(&desc, dreq->inode, FLUSH_STABLE, false,
&nfs_direct_write_completion_ops);
desc.pg_dreq = dreq;
list_for_each_entry_safe(req, tmp, &reqs, wb_list) {
/* Bump the transmission count */
req->wb_nio++;
if (!nfs_pageio_add_request(&desc, req)) {
nfs_list_move_request(req, &failed);
spin_lock(&cinfo.inode->i_lock);
dreq->flags = 0;
if (desc.pg_error < 0)
dreq->error = desc.pg_error;
else
dreq->error = -EIO;
spin_unlock(&cinfo.inode->i_lock);
}
nfs_release_request(req);
}
nfs_pageio_complete(&desc);
while (!list_empty(&failed)) {
req = nfs_list_entry(failed.next);
nfs_list_remove_request(req);
nfs_unlock_and_release_request(req);
}
if (put_dreq(dreq))
nfs_direct_write_complete(dreq);
}
static void nfs_direct_commit_complete(struct nfs_commit_data *data)
{
const struct nfs_writeverf *verf = data->res.verf;
struct nfs_direct_req *dreq = data->dreq;
struct nfs_commit_info cinfo;
struct nfs_page *req;
int status = data->task.tk_status;
if (status < 0) {
/* Errors in commit are fatal */
dreq->error = status;
dreq->max_count = 0;
dreq->count = 0;
dreq->flags = NFS_ODIRECT_DONE;
} else if (dreq->flags == NFS_ODIRECT_DONE)
status = dreq->error;
nfs_init_cinfo_from_dreq(&cinfo, dreq);
while (!list_empty(&data->pages)) {
req = nfs_list_entry(data->pages.next);
nfs_list_remove_request(req);
if (status >= 0 && !nfs_write_match_verf(verf, req)) {
dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
/*
* Despite the reboot, the write was successful,
* so reset wb_nio.
*/
req->wb_nio = 0;
nfs_mark_request_commit(req, NULL, &cinfo, 0);
} else /* Error or match */
nfs_release_request(req);
nfs_unlock_and_release_request(req);
}
if (atomic_dec_and_test(&cinfo.mds->rpcs_out))
nfs_direct_write_complete(dreq);
}
static void nfs_direct_resched_write(struct nfs_commit_info *cinfo,
struct nfs_page *req)
{
struct nfs_direct_req *dreq = cinfo->dreq;
spin_lock(&dreq->lock);
if (dreq->flags != NFS_ODIRECT_DONE)
dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
spin_unlock(&dreq->lock);
nfs_mark_request_commit(req, NULL, cinfo, 0);
}
static const struct nfs_commit_completion_ops nfs_direct_commit_completion_ops = {
.completion = nfs_direct_commit_complete,
.resched_write = nfs_direct_resched_write,
};
static void nfs_direct_commit_schedule(struct nfs_direct_req *dreq)
{
int res;
struct nfs_commit_info cinfo;
LIST_HEAD(mds_list);
nfs_init_cinfo_from_dreq(&cinfo, dreq);
nfs_scan_commit(dreq->inode, &mds_list, &cinfo);
res = nfs_generic_commit_list(dreq->inode, &mds_list, 0, &cinfo);
if (res < 0) /* res == -ENOMEM */
nfs_direct_write_reschedule(dreq);
}
static void nfs_direct_write_clear_reqs(struct nfs_direct_req *dreq)
{
struct nfs_commit_info cinfo;
struct nfs_page *req;
LIST_HEAD(reqs);
nfs_init_cinfo_from_dreq(&cinfo, dreq);
nfs_direct_write_scan_commit_list(dreq->inode, &reqs, &cinfo);
while (!list_empty(&reqs)) {
req = nfs_list_entry(reqs.next);
nfs_list_remove_request(req);
nfs_release_request(req);
nfs_unlock_and_release_request(req);
}
}
static void nfs_direct_write_schedule_work(struct work_struct *work)
{
struct nfs_direct_req *dreq = container_of(work, struct nfs_direct_req, work);
int flags = dreq->flags;
dreq->flags = 0;
switch (flags) {
case NFS_ODIRECT_DO_COMMIT:
nfs_direct_commit_schedule(dreq);
break;
case NFS_ODIRECT_RESCHED_WRITES:
nfs_direct_write_reschedule(dreq);
break;
default:
nfs_direct_write_clear_reqs(dreq);
nfs_zap_mapping(dreq->inode, dreq->inode->i_mapping);
nfs_direct_complete(dreq);
}
}
static void nfs_direct_write_complete(struct nfs_direct_req *dreq)
{
queue_work(nfsiod_workqueue, &dreq->work); /* Calls nfs_direct_write_schedule_work */
}
static void nfs_direct_write_completion(struct nfs_pgio_header *hdr)
{
struct nfs_direct_req *dreq = hdr->dreq;
struct nfs_commit_info cinfo;
bool request_commit = false;
struct nfs_page *req = nfs_list_entry(hdr->pages.next);
nfs_init_cinfo_from_dreq(&cinfo, dreq);
spin_lock(&dreq->lock);
if (test_bit(NFS_IOHDR_REDO, &hdr->flags)) {
spin_unlock(&dreq->lock);
goto out_put;
}
nfs_direct_count_bytes(dreq, hdr);
if (hdr->good_bytes != 0 && nfs_write_need_commit(hdr)) {
switch (dreq->flags) {
case 0:
dreq->flags = NFS_ODIRECT_DO_COMMIT;
request_commit = true;
break;
case NFS_ODIRECT_RESCHED_WRITES:
case NFS_ODIRECT_DO_COMMIT:
request_commit = true;
}
}
spin_unlock(&dreq->lock);
while (!list_empty(&hdr->pages)) {
nfs: add support for multiple nfs reqs per page Add "page groups" - a circular list of nfs requests (struct nfs_page) that all reference the same page. This gives nfs read and write paths the ability to account for sub-page regions independently. This somewhat follows the design of struct buffer_head's sub-page accounting. Only "head" requests are ever added/removed from the inode list in the buffered write path. "head" and "sub" requests are treated the same through the read path and the rest of the write/commit path. Requests are given an extra reference across the life of the list. Page groups are never rejoined after being split. If the read/write request fails and the client falls back to another path (ie revert to MDS in PNFS case), the already split requests are pushed through the recoalescing code again, which may split them further and then coalesce them into properly sized requests on the wire. Fragmentation shouldn't be a problem with the current design, because we flush all requests in page group when a non-contiguous request is added, so the only time resplitting should occur is on a resend of a read or write. This patch lays the groundwork for sub-page splitting, but does not actually do any splitting. For now all page groups have one request as pg_test functions don't yet split pages. There are several related patches that are needed support multiple requests per page group. Signed-off-by: Weston Andros Adamson <dros@primarydata.com> Signed-off-by: Trond Myklebust <trond.myklebust@primarydata.com>
2014-05-15 22:56:45 +07:00
req = nfs_list_entry(hdr->pages.next);
nfs_list_remove_request(req);
if (request_commit) {
kref_get(&req->wb_kref);
memcpy(&req->wb_verf, &hdr->verf.verifier,
sizeof(req->wb_verf));
nfs_mark_request_commit(req, hdr->lseg, &cinfo,
hdr->ds_commit_idx);
}
nfs_unlock_and_release_request(req);
}
out_put:
if (put_dreq(dreq))
nfs_direct_write_complete(dreq);
hdr->release(hdr);
}
static void nfs_write_sync_pgio_error(struct list_head *head, int error)
{
struct nfs_page *req;
while (!list_empty(head)) {
req = nfs_list_entry(head->next);
nfs_list_remove_request(req);
nfs_unlock_and_release_request(req);
}
}
static void nfs_direct_write_reschedule_io(struct nfs_pgio_header *hdr)
{
struct nfs_direct_req *dreq = hdr->dreq;
spin_lock(&dreq->lock);
if (dreq->error == 0) {
dreq->flags = NFS_ODIRECT_RESCHED_WRITES;
/* fake unstable write to let common nfs resend pages */
hdr->verf.committed = NFS_UNSTABLE;
hdr->good_bytes = hdr->args.offset + hdr->args.count -
hdr->io_start;
}
spin_unlock(&dreq->lock);
}
static const struct nfs_pgio_completion_ops nfs_direct_write_completion_ops = {
.error_cleanup = nfs_write_sync_pgio_error,
.init_hdr = nfs_direct_pgio_init,
.completion = nfs_direct_write_completion,
.reschedule_io = nfs_direct_write_reschedule_io,
};
/*
* NB: Return the value of the first error return code. Subsequent
* errors after the first one are ignored.
*/
/*
* For each wsize'd chunk of the user's buffer, dispatch an NFS WRITE
* operation. If nfs_writedata_alloc() or get_user_pages() fails,
* bail and stop sending more writes. Write length accounting is
* handled automatically by nfs_direct_write_result(). Otherwise, if
* no requests have been sent, just return an error.
*/
static ssize_t nfs_direct_write_schedule_iovec(struct nfs_direct_req *dreq,
struct iov_iter *iter,
loff_t pos)
{
struct nfs_pageio_descriptor desc;
struct inode *inode = dreq->inode;
ssize_t result = 0;
size_t requested_bytes = 0;
size_t wsize = max_t(size_t, NFS_SERVER(inode)->wsize, PAGE_SIZE);
nfs_pageio_init_write(&desc, inode, FLUSH_COND_STABLE, false,
&nfs_direct_write_completion_ops);
desc.pg_dreq = dreq;
get_dreq(dreq);
direct-io: only inc/dec inode->i_dio_count for file systems do_blockdev_direct_IO() increments and decrements the inode ->i_dio_count for each IO operation. It does this to protect against truncate of a file. Block devices don't need this sort of protection. For a capable multiqueue setup, this atomic int is the only shared state between applications accessing the device for O_DIRECT, and it presents a scaling wall for that. In my testing, as much as 30% of system time is spent incrementing and decrementing this value. A mixed read/write workload improved from ~2.5M IOPS to ~9.6M IOPS, with better latencies too. Before: clat percentiles (usec): | 1.00th=[ 33], 5.00th=[ 34], 10.00th=[ 34], 20.00th=[ 34], | 30.00th=[ 34], 40.00th=[ 34], 50.00th=[ 35], 60.00th=[ 35], | 70.00th=[ 35], 80.00th=[ 35], 90.00th=[ 37], 95.00th=[ 80], | 99.00th=[ 98], 99.50th=[ 151], 99.90th=[ 155], 99.95th=[ 155], | 99.99th=[ 165] After: clat percentiles (usec): | 1.00th=[ 95], 5.00th=[ 108], 10.00th=[ 129], 20.00th=[ 149], | 30.00th=[ 155], 40.00th=[ 161], 50.00th=[ 167], 60.00th=[ 171], | 70.00th=[ 177], 80.00th=[ 185], 90.00th=[ 201], 95.00th=[ 270], | 99.00th=[ 390], 99.50th=[ 398], 99.90th=[ 418], 99.95th=[ 422], | 99.99th=[ 438] In other setups, Robert Elliott reported seeing good performance improvements: https://lkml.org/lkml/2015/4/3/557 The more applications accessing the device, the worse it gets. Add a new direct-io flags, DIO_SKIP_DIO_COUNT, which tells do_blockdev_direct_IO() that it need not worry about incrementing or decrementing the inode i_dio_count for this caller. Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Elliott, Robert (Server Storage) <elliott@hp.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2015-04-16 06:05:48 +07:00
inode_dio_begin(inode);
NFS_I(inode)->write_io += iov_iter_count(iter);
while (iov_iter_count(iter)) {
struct page **pagevec;
size_t bytes;
size_t pgbase;
unsigned npages, i;
result = iov_iter_get_pages_alloc(iter, &pagevec,
wsize, &pgbase);
if (result < 0)
break;
bytes = result;
iov_iter_advance(iter, bytes);
npages = (result + pgbase + PAGE_SIZE - 1) / PAGE_SIZE;
for (i = 0; i < npages; i++) {
struct nfs_page *req;
unsigned int req_len = min_t(size_t, bytes, PAGE_SIZE - pgbase);
req = nfs_create_request(dreq->ctx, pagevec[i],
pgbase, req_len);
if (IS_ERR(req)) {
result = PTR_ERR(req);
break;
}
if (desc.pg_error < 0) {
nfs_free_request(req);
result = desc.pg_error;
break;
}
nfs_lock_request(req);
req->wb_index = pos >> PAGE_SHIFT;
req->wb_offset = pos & ~PAGE_MASK;
if (!nfs_pageio_add_request(&desc, req)) {
result = desc.pg_error;
nfs_unlock_and_release_request(req);
break;
}
pgbase = 0;
bytes -= req_len;
requested_bytes += req_len;
pos += req_len;
dreq->bytes_left -= req_len;
}
nfs_direct_release_pages(pagevec, npages);
kvfree(pagevec);
if (result < 0)
break;
}
nfs_pageio_complete(&desc);
NFS: Fix "kernel BUG at fs/aio.c:554!" Nick Piggin reports: > I'm getting use after frees in aio code in NFS > > [ 2703.396766] Call Trace: > [ 2703.396858] [<ffffffff8100b057>] ? native_sched_clock+0x27/0x80 > [ 2703.396959] [<ffffffff8108509e>] ? put_lock_stats+0xe/0x40 > [ 2703.397058] [<ffffffff81088348>] ? lock_release_holdtime+0xa8/0x140 > [ 2703.397159] [<ffffffff8108a2a5>] lock_acquire+0x95/0x1b0 > [ 2703.397260] [<ffffffff811627db>] ? aio_put_req+0x2b/0x60 > [ 2703.397361] [<ffffffff81039701>] ? get_parent_ip+0x11/0x50 > [ 2703.397464] [<ffffffff81612a31>] _raw_spin_lock_irq+0x41/0x80 > [ 2703.397564] [<ffffffff811627db>] ? aio_put_req+0x2b/0x60 > [ 2703.397662] [<ffffffff811627db>] aio_put_req+0x2b/0x60 > [ 2703.397761] [<ffffffff811647fe>] do_io_submit+0x2be/0x7c0 > [ 2703.397895] [<ffffffff81164d0b>] sys_io_submit+0xb/0x10 > [ 2703.397995] [<ffffffff8100307b>] system_call_fastpath+0x16/0x1b > > Adding some tracing, it is due to nfs completing the request then > returning something other than -EIOCBQUEUED, so aio.c > also completes the request. To address this, prevent the NFS direct I/O engine from completing async iocbs when the forward path returns an error without starting any I/O. This fix appears to survive ^C during both "xfstest no. 208" and "fsx -Z." It's likely this bug has existed for a very long while, as we are seeing very similar symptoms in OEL 5. Copying stable. Cc: Stable <stable@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-01-21 22:54:57 +07:00
/*
* If no bytes were started, return the error, and let the
* generic layer handle the completion.
*/
if (requested_bytes == 0) {
direct-io: only inc/dec inode->i_dio_count for file systems do_blockdev_direct_IO() increments and decrements the inode ->i_dio_count for each IO operation. It does this to protect against truncate of a file. Block devices don't need this sort of protection. For a capable multiqueue setup, this atomic int is the only shared state between applications accessing the device for O_DIRECT, and it presents a scaling wall for that. In my testing, as much as 30% of system time is spent incrementing and decrementing this value. A mixed read/write workload improved from ~2.5M IOPS to ~9.6M IOPS, with better latencies too. Before: clat percentiles (usec): | 1.00th=[ 33], 5.00th=[ 34], 10.00th=[ 34], 20.00th=[ 34], | 30.00th=[ 34], 40.00th=[ 34], 50.00th=[ 35], 60.00th=[ 35], | 70.00th=[ 35], 80.00th=[ 35], 90.00th=[ 37], 95.00th=[ 80], | 99.00th=[ 98], 99.50th=[ 151], 99.90th=[ 155], 99.95th=[ 155], | 99.99th=[ 165] After: clat percentiles (usec): | 1.00th=[ 95], 5.00th=[ 108], 10.00th=[ 129], 20.00th=[ 149], | 30.00th=[ 155], 40.00th=[ 161], 50.00th=[ 167], 60.00th=[ 171], | 70.00th=[ 177], 80.00th=[ 185], 90.00th=[ 201], 95.00th=[ 270], | 99.00th=[ 390], 99.50th=[ 398], 99.90th=[ 418], 99.95th=[ 422], | 99.99th=[ 438] In other setups, Robert Elliott reported seeing good performance improvements: https://lkml.org/lkml/2015/4/3/557 The more applications accessing the device, the worse it gets. Add a new direct-io flags, DIO_SKIP_DIO_COUNT, which tells do_blockdev_direct_IO() that it need not worry about incrementing or decrementing the inode i_dio_count for this caller. Cc: Andrew Morton <akpm@linux-foundation.org> Cc: Christoph Hellwig <hch@lst.de> Cc: Theodore Ts'o <tytso@mit.edu> Cc: Elliott, Robert (Server Storage) <elliott@hp.com> Cc: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Jens Axboe <axboe@fb.com> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
2015-04-16 06:05:48 +07:00
inode_dio_end(inode);
NFS: Fix "kernel BUG at fs/aio.c:554!" Nick Piggin reports: > I'm getting use after frees in aio code in NFS > > [ 2703.396766] Call Trace: > [ 2703.396858] [<ffffffff8100b057>] ? native_sched_clock+0x27/0x80 > [ 2703.396959] [<ffffffff8108509e>] ? put_lock_stats+0xe/0x40 > [ 2703.397058] [<ffffffff81088348>] ? lock_release_holdtime+0xa8/0x140 > [ 2703.397159] [<ffffffff8108a2a5>] lock_acquire+0x95/0x1b0 > [ 2703.397260] [<ffffffff811627db>] ? aio_put_req+0x2b/0x60 > [ 2703.397361] [<ffffffff81039701>] ? get_parent_ip+0x11/0x50 > [ 2703.397464] [<ffffffff81612a31>] _raw_spin_lock_irq+0x41/0x80 > [ 2703.397564] [<ffffffff811627db>] ? aio_put_req+0x2b/0x60 > [ 2703.397662] [<ffffffff811627db>] aio_put_req+0x2b/0x60 > [ 2703.397761] [<ffffffff811647fe>] do_io_submit+0x2be/0x7c0 > [ 2703.397895] [<ffffffff81164d0b>] sys_io_submit+0xb/0x10 > [ 2703.397995] [<ffffffff8100307b>] system_call_fastpath+0x16/0x1b > > Adding some tracing, it is due to nfs completing the request then > returning something other than -EIOCBQUEUED, so aio.c > also completes the request. To address this, prevent the NFS direct I/O engine from completing async iocbs when the forward path returns an error without starting any I/O. This fix appears to survive ^C during both "xfstest no. 208" and "fsx -Z." It's likely this bug has existed for a very long while, as we are seeing very similar symptoms in OEL 5. Copying stable. Cc: Stable <stable@kernel.org> Signed-off-by: Chuck Lever <chuck.lever@oracle.com> Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>
2011-01-21 22:54:57 +07:00
nfs_direct_req_release(dreq);
return result < 0 ? result : -EIO;
}
if (put_dreq(dreq))
nfs_direct_write_complete(dreq);
return requested_bytes;
}
/**
* nfs_file_direct_write - file direct write operation for NFS files
* @iocb: target I/O control block
* @iter: vector of user buffers from which to write data
*
* We use this function for direct writes instead of calling
* generic_file_aio_write() in order to avoid taking the inode
* semaphore and updating the i_size. The NFS server will set
* the new i_size and this client must read the updated size
* back into its cache. We let the server do generic write
* parameter checking and report problems.
*
* We eliminate local atime updates, see direct read above.
*
* We avoid unnecessary page cache invalidations for normal cached
* readers of this file.
*
* Note that O_APPEND is not supported for NFS direct writes, as there
* is no atomic O_APPEND write facility in the NFS protocol.
*/
ssize_t nfs_file_direct_write(struct kiocb *iocb, struct iov_iter *iter)
{
ssize_t result = -EINVAL, requested;
size_t count;
struct file *file = iocb->ki_filp;
struct address_space *mapping = file->f_mapping;
struct inode *inode = mapping->host;
struct nfs_direct_req *dreq;
struct nfs_lock_context *l_ctx;
loff_t pos, end;
dfprintk(FILE, "NFS: direct write(%pD2, %zd@%Ld)\n",
file, iov_iter_count(iter), (long long) iocb->ki_pos);
result = generic_write_checks(iocb, iter);
if (result <= 0)
return result;
count = result;
nfs_add_stats(mapping->host, NFSIOS_DIRECTWRITTENBYTES, count);
pos = iocb->ki_pos;
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
end = (pos + iov_iter_count(iter) - 1) >> PAGE_SHIFT;
task_io_account_write(count);
result = -ENOMEM;
dreq = nfs_direct_req_alloc();
if (!dreq)
goto out;
dreq->inode = inode;
dreq->bytes_left = dreq->max_count = count;
dreq->io_start = pos;
dreq->ctx = get_nfs_open_context(nfs_file_open_context(iocb->ki_filp));
l_ctx = nfs_get_lock_context(dreq->ctx);
if (IS_ERR(l_ctx)) {
result = PTR_ERR(l_ctx);
nfs_direct_req_release(dreq);
goto out_release;
}
dreq->l_ctx = l_ctx;
if (!is_sync_kiocb(iocb))
dreq->iocb = iocb;
pnfs_init_ds_commit_info_ops(&dreq->ds_cinfo, inode);
nfs_start_io_direct(inode);
requested = nfs_direct_write_schedule_iovec(dreq, iter, pos);
if (mapping->nrpages) {
invalidate_inode_pages2_range(mapping,
mm, fs: get rid of PAGE_CACHE_* and page_cache_{get,release} macros PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} macros were introduced *long* time ago with promise that one day it will be possible to implement page cache with bigger chunks than PAGE_SIZE. This promise never materialized. And unlikely will. We have many places where PAGE_CACHE_SIZE assumed to be equal to PAGE_SIZE. And it's constant source of confusion on whether PAGE_CACHE_* or PAGE_* constant should be used in a particular case, especially on the border between fs and mm. Global switching to PAGE_CACHE_SIZE != PAGE_SIZE would cause to much breakage to be doable. Let's stop pretending that pages in page cache are special. They are not. The changes are pretty straight-forward: - <foo> << (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - <foo> >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) -> <foo>; - PAGE_CACHE_{SIZE,SHIFT,MASK,ALIGN} -> PAGE_{SIZE,SHIFT,MASK,ALIGN}; - page_cache_get() -> get_page(); - page_cache_release() -> put_page(); This patch contains automated changes generated with coccinelle using script below. For some reason, coccinelle doesn't patch header files. I've called spatch for them manually. The only adjustment after coccinelle is revert of changes to PAGE_CAHCE_ALIGN definition: we are going to drop it later. There are few places in the code where coccinelle didn't reach. I'll fix them manually in a separate patch. Comments and documentation also will be addressed with the separate patch. virtual patch @@ expression E; @@ - E << (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ expression E; @@ - E >> (PAGE_CACHE_SHIFT - PAGE_SHIFT) + E @@ @@ - PAGE_CACHE_SHIFT + PAGE_SHIFT @@ @@ - PAGE_CACHE_SIZE + PAGE_SIZE @@ @@ - PAGE_CACHE_MASK + PAGE_MASK @@ expression E; @@ - PAGE_CACHE_ALIGN(E) + PAGE_ALIGN(E) @@ expression E; @@ - page_cache_get(E) + get_page(E) @@ expression E; @@ - page_cache_release(E) + put_page(E) Signed-off-by: Kirill A. Shutemov <kirill.shutemov@linux.intel.com> Acked-by: Michal Hocko <mhocko@suse.com> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2016-04-01 19:29:47 +07:00
pos >> PAGE_SHIFT, end);
}
nfs_end_io_direct(inode);
if (requested > 0) {
result = nfs_direct_wait(dreq);
if (result > 0) {
requested -= result;
iocb->ki_pos = pos + result;
/* XXX: should check the generic_write_sync retval */
generic_write_sync(iocb, result);
}
iov_iter_revert(iter, requested);
} else {
result = requested;
}
out_release:
nfs_direct_req_release(dreq);
out:
return result;
}
/**
* nfs_init_directcache - create a slab cache for nfs_direct_req structures
*
*/
int __init nfs_init_directcache(void)
{
nfs_direct_cachep = kmem_cache_create("nfs_direct_cache",
sizeof(struct nfs_direct_req),
0, (SLAB_RECLAIM_ACCOUNT|
SLAB_MEM_SPREAD),
NULL);
if (nfs_direct_cachep == NULL)
return -ENOMEM;
return 0;
}
/**
* nfs_destroy_directcache - destroy the slab cache for nfs_direct_req structures
*
*/
void nfs_destroy_directcache(void)
{
kmem_cache_destroy(nfs_direct_cachep);
}